Evaluating the consistency, repeatability, and reproducibility of osteometric data on dry bone surfaces, scanned dry bone surfaces, and scanned bone surfaces obtained from living individuals

Applying dry bone osteometrics to virtual bone surfaces obtained via medical imaging raises the question of consistency between the variables. Variables obtained from virtual bone surfaces also need to be sufficiently repeatable and reproducible to be valid for anthropological studies. This is also true for the landmarks defining these variables and for their acquisition. The consistency between variables taken directly from dry bones and from the virtual surfaces of dry bones was tested on 40 clavicles. 30 virtual surfaces of iliae, fifth lumbar vertebrae, and clavicles reconstructed from computed tomography scans of living individuals were used to test the repeatability and reproducibility of 16 landmarks and 19 variables. Statistical tests, graphical and quantitative error evaluations, and intraclass correlation coefficients were applied. The differences between all variables taken on dry and virtual clavicles were less than ±1 mm. Bland–Altman plots showed more than 95% reliability between variables obtained on dry bone and their virtually reconstructed surfaces, confirming their consistency and thus validating their use in osteometric studies independently of the medium of study. Although not all landmarks were repeatable and reproducible, most variables were. To assess intra- or inter-observer errors, graphical representations or coefficients are more precise and accurate than statistical tests. These two evaluation methods should be given priority to test the repeatability and reproducibility of osteometric variables.